Arxi
Programs log; agents don't—not with the integrity that audits, replays, and compliance demand. Arxi restores that integrity with append-only, hash-chained evidence.
Recording What Actually Happened
Agents call tools, pipelines fire, sensors report—but the boundary events multiply and scatter. Most systems capture these as best-effort text logs. The evidence isn’t chained, isn’t verifiable, and can’t be replayed independently.
If you can’t prove what happened at a boundary, you can’t prove your system works.
Arxi closes that gap. It captures boundary events as cryptographically chained evidence envelopes—append-only, tamper-evident, and verifiable offline. Any process that crosses a trust boundary can be recorded.
The same automation becomes auditable in places it couldn’t be before—because now there’s a verifiable chain of what happened and why.
What you get
- Tamper-evident recording (append-only, hash-chained envelopes)
- Verifiable evidence bundles (exportable, independently verifiable offline)
- Deterministic integrity (same inputs produce identical hashes via JCS canonical encoding)
What Arxi does not do It doesn’t make decisions or enforce policies. It records boundary events with integrity—so decision systems, auditors, and humans have proof to work with.
Where it shows up A few common examples:
- Agent and tool execution trails
- CI/CD pipeline evidence capture
- Compliance and regulatory audit records
Where to go next Start with Basics for the core ideas, then Examples for workflows, then Docs for the full reference.
- Arxi Basics — The three core concepts (envelopes, chains, bundles)
- Arxi Examples — Three real workflows, no jargon
- Full Documentation — Specs, tooling, and reference